Digital image edge detection and road network tracking method and system

Abstract

A system and method for finding edge orientation and magnitude, and extracting a road path from an image. An image is segmented into at least a first window and a second window. A first road segment having plural parameters within the first window is specified. Plural road segments are identified, each having plural parameters, in the second window. The orientation of plural line segments in the second window are determined, and paired, based on their orientation. The line segment orientations are determined by receiving a plurality of edge orientation estimates, transforming a portion of them, and aggregating the transformed portion into an edge orientation estimate. A stochastic process is performed on the plural parameters of the plural road segments to identify a second road segment having the maximum correlation with the plural parameters of the first road segment.

Description

[0001] 1. Field of the Invention[0002] The present invention relates to digital image analysis and mapping. More specifically, the present invention relates to edge detection and road network tracking in aerial digital images.[0003] 2. Description of the Related Art[0004] Aerial images are produced by cameras located in aircraft or other flying machines and from orbiting satellites. Such cameras are typically pointed downward and are used to capture images of the Earth's surface. While photosensitive film and chemical processing have been employed to capture images in the past, modern aerial imaging systems typically employ digital imaging sensors that output digital image files. The digital image files produced by digital imaging sensors can be transmitted by radio communications or through transportation of digital data storage media. The file structure typically includes a plurality of pixels that are located in two dimensions and that comprise data reflecting the electromagnetic...

AI Technical Summary

Problems solved by technology

989. Orientation estimates obtained by using compass type directional filters are inherently inaccurate, due to the quantization of the orientation values into ranges specified by the pixel map size and number of directional filter responses empl

The prior art methods of doing this do not handle the multi-spectral images, or require segmentation-based highly complex operations in order to calculate line / edge strengths.

Despite the number of different approaches that have been proposed for edge and line detection, the accurate estimation of the edge orientation has only been marginally investigated.

Most methods such as derivative approaches using linear filtering, mathematical morphology, Markov fields, and surface models, concentrate only on the accurate localization of the edge and its immunity to noise, but not the precise estimation of the edge orientation.

Part of the problem with the prior art is that it was designed for a single energy band image.

With higher resolution data, there are new challenges, such as shadows across the roadway, trees on a road edge, cars, so it is more complex.

Also, the prior art Markov chain did not adapt well to changes in road surface color intensity (asphalt to concrete, etc.).

Although some prior art techniques that have been developed to extract road networks from aerial imagery may be modified to deal with high resolution satellite imagery, the road networks present in high resolution satellite images are much more complex.

This has created an environment where the prior art algorithms are unable to extract all the useful information available in the higher resolution imagery.

As mentioned above, high-resolution imagery road networks present more details on the roads such as lane markings, vehicles, shadows cast by buildings and trees, overpasses and changes in surface material, which make the extraction of road networks a much more complicated problem.

As a result, most existing algorithms are not suitable to process high-resolution imagery.

In addition, most existing algorithms are deterministic and model-free techniques, which is another reason why they can't handle the variations of road networks presented in high resolution-imagery.

707-721, 1996, Barzohar and Cooper proposed a geometric-stochastic model of roads for finding main road networks, but this model is not optimal for two reasons.

These variations make the extraction of road networks a much more complicated problem.

Images